nipalsMCIA

nipalsMCIA implements Multiple Co-Inertia Analysis (MCIA), a joint dimensionality reduction (jDR) algorithm, for multi-block datasets using an extension to the Nonlinear Iterative Partial Least Squares (NIPALS) method to enable unsupervised learning for clustering, visualization, and feature selection in multi-omics analyses.

Key Features:

• NIPALS-based MCIA computation: Solves the MCIA objective function using an extension to the Nonlinear Iterative Partial Least Squares method as proposed by Hanafi et al. (2010).
• Flexible preprocessing: Provides multiple options for row-level and table-level preprocessing of multi-block data.
• Accelerated variance computation: Speeds up computation of variance explained compared to other implementations.
• Scalability: Achieves significant speed-up over other implementations for datasets with large sample size and/or feature dimension.

Scientific Applications:

• Bulk multi-omics analysis: Enables joint dimensionality reduction, clustering, visualization, and feature selection for bulk multi-omics datasets.
• Single-cell multi-omics analysis: Supports unsupervised learning for clustering, visualization, and feature selection in single-cell multi-omics studies.

Methodology:

The method extends the Nonlinear Iterative Partial Least Squares (NIPALS) algorithm to solve the Multiple Co-Inertia Analysis objective function for joint dimensionality reduction of multi-block datasets.